One hundred patients with rectal adenocarcinoma were examined preoperatively with rectal endosonography (ES) and 50 were also examined with computed tomography (CT). ES predicted mesorectal lymph node involvement with an accuracy of 83 per cent, sensitivity of 88 per cent, specificity of 79 per cent, positive predictive value of 78 per cent and negative predictive value of 89 per cent. CT in comparison had an accuracy of 57 per cent, sensitivity of 25 per cent, specificity of 91 per cent, positive predictive value of 75 per cent and negative predictive value of 53 per cent. No particular histological architectural feature could be identified as responsible for false positive diagnosis though nodal size was significantly larger in the true positive and false positive group when compared with the true negatives (P less than 0.001 and P less than 0.01 respectively).

Forty-four patients with primary rectal cancers and six patients with benign rectal lesions were examined pre-operatively digitally, with endorectal sonography (ELU) and also computed tomography (CT). Digital examination of the rectal cancers had an overall accuracy of 68 per cent and predicted invasion beyond the muscularis propria with a sensitivity of 68 per cent, specificity of 83 per cent, positive predictive value of 100 per cent and negative predictive value of 46 per cent. In comparison CT had an accuracy of 82 per cent, sensitivity of 86 per cent, specificity of 62 per cent, positive predictive value of 91 per cent and negative predictive value of 50 per cent. ELU was the most reliable indicator of local invasion in rectal cancer when compared with postoperative histopathology with an accuracy of 91 per cent, sensitivity of 94 per cent, specificity of 87 per cent, positive predictive value of 97 per cent and negative predictive value of 78 per cent.

With the increasing use of transrectal sonography, accurate preoperative staging of rectal cancer requires correct identification of the normal ultrasonographic appearances of the colon and rectum. Fifteen rectal and colonic specimens have been studied in vitro to define the normal anatomy. Five distinct ultrasonic layers have been identified; a first echogenic layer that corresponds to the mucosa, a second echopoor layer made up of mucosa and muscularis mucosae, a third echogenic layer that is submucosa, a fourth echopoor layer that is muscularis propria, and a fifth echogenic layer made up of serosa and perirectal fat.

Eighty-five patients treated surgically for rectal cancer have been followed up by conventional clinical examination, sigmoidoscopy, and endosonography. Local recurrence was diagnosed in 22 patients. Nineteen of these had either sigmoidoscopic or digital evidence of recurrence and three were diagnosed solely by endosonography. In all cases endosonography gave additional information on which to base management decisions. Routine use of endosonography should allow the detection of early recurrence in a larger number of patients.